At a glance. Acme's Q1 2026 investor conference — Taiwan's equivalent of a US earnings call — turned a year-earlier loss into EPS of NT$0.1 on AI-driven ferrite demand. The 800V item is already two-legged: soft ferrite cores are shipping into HVDC 800V power supplies, and N-type SiC powder is in sampling for AI-server 800V PSUs. Neither share of sales is a number.
What changed this quarter
Ferrite ran on AI servers and data centers. That is what lifted revenue 18.6% and gross profit 52%. Notebooks and games consoles were flat; automotive electronics were slightly down on uneven EV demand in Europe and the US.
The company returned to profit after a NT$31m net loss a year earlier. Underneath that, Guangzhou's high-tech-enterprise status was revoked, triggering back taxes and late fees of about NT$13.4m — a one-off below the line, not a product miss.
SiC remained a sampling and mix story, not a volume story. N-type powder capacity is built; semi-insulating capacity is the line being added. New materials are 16–18% of sales.
The numbers
| Q1 2026 | Value | Change |
|---|---|---|
| Revenue | NT$853m (~US$27m) | +18.6% YoY; slightly below 2025Q4 on the Lunar New Year holiday |
| Gross profit | NT$166m (~US$5.3m) | +52% YoY |
| Operating profit | NT$25m (~US$0.8m) | — |
| Net income | NT$13m (~US$0.4m) | vs a NT$31m loss in 2025Q1 |
| Net to parent | NT$22m (~US$0.7m) | — |
| EPS | NT$0.1 (~US$0.003) | — |
The P&L summary showed a NT$9m non-operating net loss. In Q&A, management described a NT$14.8m non-operating net loss, of which about NT$13.4m was the Guangzhou tax restatement. Both figures are as stated; they were not reconciled on the call.
Debt-to-assets 51%. Long-term borrowings about NT$1.85bn (~US$59m), costing roughly NT$11m of interest in the quarter. Depreciation is already above NT$100m a quarter, all of it on equipment that is running — management said there are no idle assets. No capital raise or bond issue is planned.
NVIDIA 800V read-through
Two materials legs, at different stages of commercial life.
Ferrite cores are already in supply for AI servers and for HVDC 800V power architectures. Asked directly whether high-frequency magnetics had been designed into those platforms, management said yes, they are shipping. AI-server grades are specified more tightly than commodity cores; the visible step is small, the ASP and margin are not. Some "star" grades run gross margin above 30%. The AI-data-center share of group sales is rising, but customers do not always name the end use, so there is no mix percentage.
N-type SiC powder is the other leg: crystal-growth feedstock for power devices that would sit in an AI-server / data-center HVDC 800V PSU, on the back of high thermal conductivity and high-voltage rating. It is in sampling or customer discussion, not a finished-device shipment. Semi-insulating SiC is a different application (higher-frequency RF), also in sampling. Asked whether both businesses show up in data-center HVDC, management said both units have product in that architecture.
This is a materials slot — magnetics in the converter, SiC powder upstream of the switch — not a PSU or a rack. There is no customer name, no wattage, and no 800V revenue figure.
Segments that moved
No numbered split between ferrite and SiC was given beyond new businesses at 16–18% of sales.
Ferrite is the cash cow and the AI print. Utilisation about 90%. Demand is up and the company said it can still deliver, with some local equipment adds. TrendForce figures cited on the call: 2026 AI-server shipments up more than 28%, total servers +12.8%.
SiC. Utilisation about 50–60%. Q1 N-type volumes were higher, Q2 a bit lower — described as short-term customer inventory, with long-term usage still rising. Semi-insulating demand is the tight one (defence, space, high-frequency RF); that capacity has been the recent build. China spot prices for SiC are already very low; Acme's answer is non-China customers plus higher purity, and it said it has not been obviously hurt. High-purity powder has had no customer complaints to date. AlN is developed and in customer qualification.
Guidance
- Gross margin still has room from mix, utilisation and high-end penetration — described as an ongoing process, not a target number.
- Ferrite demand from AI is the clear near-term; N-type SiC is lumpy near term and up long term; semi-insulating SiC is the capacity constraint.
- Semi-insulating SiC powder and AlN are the expansion priorities; downstream bulk and light machining will be considered if customers need it.
- No group revenue or EPS guide.
Management Q&A
Q: Quarterly depreciation is already above NT$100m; will it keep rising, and what revenue is needed to break even?
A: The investment is for future growth, mainly the new businesses, and a prepaid-equipment balance of NT$113m will still move into fixed assets. Every tool is already running, with no idle assets, so the depreciation is treated as a positive. Breakeven depends on product mix rather than a single revenue number; what matters is whether the invested equipment is utilised.
Q: Long-term borrowings have reached NT$1.85bn and interest costs are high; are there plans to refinance at lower rates or to raise capital?
A: Financial policy is conservative, bank relationships are good, and both facility size and tenor are described as favourable, with quarterly finance cost about NT$11m, so there is no funding gap. There is no plan to raise equity or issue bonds now. A large future capex need would not rule out an equity raise.
Q: What caused the NT$14.8m net non-operating loss in Q1 2026, and could it recur?
A: Most of it came from the Guangzhou subsidiary losing high-tech-enterprise status after a stricter local audit, which triggered back VAT, urban-construction tax and late fees of about NT$13.4m. The company will re-apply next year and prepare a tighter file so a similar cancellation is less likely.
Q: What is the progress of high-purity SiC powder capacity?
A: N-type capacity build-out is complete. Recent work has been on semi-insulating capacity, because defence, space and other high-frequency demand has risen sharply, and that build-out has also been largely completed.
Q: What share of revenue is new-product SiC, and how do scale economies look?
A: SiC is about 16–18% of revenue. Scale is not the main issue; timing of the investment and utilisation of the tools matter more, and both mix and utilisation are improving.
Q: Can silicon carbide enter AI applications?
A: Yes. N-type power devices can go into AI-server and data-center HVDC 800V supplies because of heat dissipation and high-voltage capability, and semi-insulating material can be used at still higher frequencies. Both are already in sampling or customer discussions, though the company cannot always confirm the customer's end use.
Q: There is a rumour that TSMC CoWoS uses sapphire; will Acme try to become a supplier?
A: Acme is treating the rumour calmly and focusing on promoting its materials technology. It sees a chance to enter the chain directly or indirectly, but will not use a named customer's brand in promotion.
Q: Have high-frequency magnetic materials already been designed into AI servers or HVDC 800V power architectures?
A: They are already shipping into those applications.
Q: How different are AI-server magnetic-material specs versus traditional parts?
A: The requirements are much finer, like chasing a world record: the step-up is small, but the characteristic difference is what matters. Unit price and gross margin are better, and that is where R&D is being directed.
Q: Where is high-purity SiC powder used, and what are qualification status and capacity?
A: Uses include conductive (N-type), semi-insulating, and semiconductor-equipment consumables, and many grades have already been qualified and are shipping. N-type capacity is 20 tonnes a year; semi-insulating is 1 tonne and still being raised.
Q: How intense is Chinese SiC price competition, and what is Acme's edge?
A: Chinese market prices are already very low. Acme's positioning is to lock onto non-China customers — a geopolitical split — and to differentiate on purity and characteristics, so it has not been clearly hurt by Chinese price competition.
Q: Why have gross margins improved in recent years, and is there more room?
A: The three drivers are mix optimisation, higher utilisation, and more high-end product penetration. Management still sees room to keep improving.
Q: Over the next three years, which core lines should the market re-learn?
A: Ferrite cores are the cash-cow that funds new work and are being steered toward higher-value, AI-related parts. In new businesses, high-purity SiC powder is the fastest, semi-insulating has been the standout recently, and aluminium nitride has finished development and is in customer qualification.
Q: What is the customer base, and what is the expansion plan?
A: Ferrite customers are fairly balanced geographically, with Taiwan companies the largest slice; new-business customers are global. Near-term expansion is mainly semi-insulating SiC powder and aluminium nitride, with downstream processing considered if customers need it; N-type SiC capacity is already seen as sufficient.
Q: As AI power architectures move to higher voltage, frequency and efficiency, what is Acme's role in the supply chain?
A: The company is aiming at the parts of that stack that need high frequency, heat resistance, heat dissipation, high efficiency and harsher operating environments, which it says it already has in hand.
Q: Will rising oil prices lift electric-vehicle and silicon-carbide demand?
A: EV demand has increased in Europe, China and the United States, but whether that is lasting still has to be watched. Any lift also depends on customer inventory policy, so it will not show through immediately.
Q: What is the company outlook?
A: R&D is being focused on future uses, especially AI, and also robotics, autonomous driving and broader changes in daily life. Customers are managed on a global, not one-or-two-account, basis, and operations are being tightened without cutting technology, quality or customer commitments.
Q: How are the two main businesses used in data-center HVDC architectures?
A: Products from both businesses are already being used in that architecture.
Q: Can ferrite-core capacity keep up with AI demand, and how is delivery?
A: Demand is higher, operations are on track, some equipment has been added locally, and deliveries are being met, including the incremental volume.
Q: How is demand for high-purity SiC powder, and have AI data centers stretched lead times?
A: Q1 volume was higher and Q2 a bit lower, which is a short inventory swing; customers' long-term usage is still rising. Lead times are still acceptable because capacity was built ahead, though the semi-insulating slice is somewhat tight.
Q: What is utilisation in the two businesses?
A: Ferrite cores are about 90%. Silicon carbide is about 50–60%.
Q: Is gross margin on data-center ferrite above 30%?
A: An average is hard to state because mix moves around. Some star products are above 30%.
Q: With the HVDC trend and capacity expansion complete, is the company in a favourable position?
A: Management compared it to a changeable market: if technology and operating capability are in place, the company should keep improving, and it is working to sustain growth.
Q: What share of revenue is AI data centers?
A: Customers do not always disclose the end use, so a precise share is hard to tally. The share has been rising.
Q: Where is the production barrier in high-purity SiC powder?
A: The barrier is excellent operations plus technology — process know-how, scale, production stability and quality; Acme says it has had no product customer complaints, only occasional packaging issues. Wolfspeed makes some powder in-house on crystal-growth tools and still buys some externally.
Disclaimer
English notes on Acme Electronics' investor conference of 20 May 2026, covering Q1 2026. Figures and statements are as presented by management and have not been independently verified. USD equivalents are approximate at about NT$31.5 to the dollar. For reference only; not investment advice.